A thrust control method of variable cycle aeroengine based on q-learning

Abstract

The invention provides a Q-Learning-based variable-cycle aero-engine thrust control method, which belongs to the field of aero-engine control and simulation technology. parameters, combined with the ε-greedy strategy, to obtain the appropriate fuel flow of the variable cycle aeroengine; further, control the thrust of the variable cycle aeroengine according to the fuel flow, and update the action value function according to the feedback of the system. The invention adopts the Q-Learning method to construct the thrust controller of the variable-cycle aeroengine, and with the increase of training times, the dynamic and steady-state characteristics of the engine are gradually improved, and the engine performance can be significantly improved. In addition, the Q‑Learning controller can continuously accumulate experience and quickly adjust the fuel flow for different control commands or thrust requirements, so as to realize rapid and stable control of variable cycle aeroengines with self-learning ability.

Description

A variable-cycle aero-engine thrust control method based on Q-Learning technical field The invention belongs to aero-engine control and simulation technical field, be specifically related to a kind of based on Q-Learning Variable-cycle aero-engine thrust control method. Background technique The variable cycle aero-engine is changed by changing the size, shape and position of some variable geometry components of the engine. The cycle parameters of the engine improve the overall performance of the engine propulsion system. At the same time, the variable cycle aero-engine has a variety of Working mode, it can show good performance under subsonic, transonic, supersonic and hypersonic conditions. but a quick A fast, accurate, stable and self-learning control system is the precondition to ensure the overall performance of the engine. Traditional control methods such as PID control, active disturbance rejection control, fuzzy control, etc. It is difficult for complex ...

AI Technical Summary

Problems solved by technology

However, a fast, accurate, stable control system with a certain self-learning ability is a prerequisite for ensuring the overall performance of the engine, and traditional control methods such as PID control, active disturbance rejection control, and fuzzy control have great potential for variable-cycle aeroengines. For complex systems with strong nonlinearity, multi-variables, and multi-work modes, it is difficult to achieve ideal control performance with self-learning ability under full-envelope and variable working conditions

In addition, the traditional control technology is difficult to meet the control accuracy requirements for complex systems, poor robustness for strong nonlinear systems, and slow response speed.

Images